This announces the availability of version 1.2.13pl8 of Linux/68k. The source and a precompiled kernel are in /pub/Linux/LOCAL/680x0 on ftp.uni-erlangen.de and its mirrors. Changes against the last version: - Loadable console fonts are now implemented. They work by new ioctls PIO_XFONT and GIO_XFONT, that know about font sizes (in opposite to PCish [PG]IO_FONT). For more details to the ioctls, see Documentation/m68k/differences.txt. BTW, every console can have a different font. A utility 'setfont' to load console fonts can be found in the bin/ directory of uni-erlangen. (myself) - Following an idea by Joerg Dorchain, there's now a version checking between the bootstrap and the kernel. This should avoid some user confusion in future. It works as follows: The bootinfo interface between bootstrap and kernel is assigned a version number, different for each machine type. The bootstrap checks the version in the kernel against its own version, and refuses to boot if the major numbers are different. If only minors differ, bootstrap gives a warning if the kernel is newer. Current bootinfo versions start from 1.0 for both, Amiga and Atari. (myself) - Some fixes to the Amiga floppy driver. Mounted filesystems should work now. Also fixed the exported device size for 'tar -M'. (dorchain@cscip.uni-sb.de) - Integrated the driver for the BioNet-100 Atari Ethernet adaptor (manufactured by Biodata GmbH, Lichtenfels, Germany). That adaptor is attached to the ACSI bus, so you need ACSI support in the kernel. The driver itself can also be built as a module. (laue@ifk-mp.uni-kiel.de, narjes@ifk-mp.uni-kiel.de) In difference to the original version of the driver, I made it compile also as integrate part of the kernel, not only as a module. Additionally, the device name is always "eth?", not "bio0" anymore if loaded as module. To avoid code redundancy, I also made some more functions of acsi.c public and gave them a "acsi_" prefix. Another point: Seemed that the BioNet driver sometimes left the ACSI/Floppy interrupt disabled... (myself) - Removed a NULL pointer access in the MFP serial driver. (dorchain@cscip.uni-sb.de) - Fixed another probing bug of the Atari Lance driver: On the Falcon with its 24-bit bus, testing 0xfe010000 works at phys. 0x00010000, where kernel code resides. Writing to memory around there caused bad crashes... The fix is to restore changed values at the probed addresses. (myself, with help by Michael Schmitz) - Big patch packet from Andreas : o Implemented the getdents() system call for GNU libc (from 1.3). o Moved system call number definition into include/asm. o Prepare bootstraps for new linker that puts the header into the first segment. o Set blk_sizes[] for ramdisk o Avoid empty for loops in some Makefiles o Increase max. process size to 3.75 GB; shared memory now at 0xc0000000 o Replace some constants by symbols o some minor fixes from 1.3 - Remvoed the m68k #ifdef's in sched.c by 1) introducing a instruction_pointer() macro abstraction as in 1.3 and 2) using an alternate version of user_mode() in arch-dependent code instead of in do_timer(). (myself) - The wd33c93 driver now doesn't use synchronous transfers (on any target) by default, since that often caused trouble. If you wish sync transfers, use the "wd33c93=" command line option. - Fixed reentrancy problem of the Amiga builtin serial driver. Also the TX int handler has been removed from the machspec list. Instead it's called directly, since no other ints is on that level. Another change: now 'amiga_colorclock' is returned for the baud base. (Matthias.Welwarsky@ppp.th-darmstadt.de) - Fixed crashes of the GVP-II SCSI driver if falling back to PIO on scatter-gather transfers. Chipmem bounce buffers are now used in that case. (Matthias.Welwarsky@ppp.th-darmstadt.de) - The following Atari drivers can now be built as module: o Lance Ethernet o MFP serial ports o SCC o MIDI This modularization required 1) that all these drivers are selectable by 'make config' (I did this for Amiga-builtin also) and 2) changing the initialization scheme of serial ports. Basics: Each driver's init function is called unconditionally and it should --if the appropriate hardware is found-- request serial slots with register_serial(). I also had to change the Amiga driver's initialization, I hope I didn't break anything. BTW, the IO-Extender and MFP-III drivers seem to rely on consecutive line numbers. This is still true for now, but not for all time... (myself) Current Amiga bootstrap version: 3.1 (changed) Current Amiga-Lilo version : 0.7 (changed) Current Atari bootstrap version: 1.7 (changed) The precompiled kernel contains both, Amiga and Atari support, so it is very big. You probably want to compile your own kernel tailored to your personal needs. Roman